AI: More than Human (Barbican Curve Gallery, London, 16 May - 25 August 2019)
Reviewed by Hallvard Haug (Birkbeck College London)
In recent years, artificial intelligence has emerged yet again from hibernation to dominate the technological sphere. The commentariat frets over how machine-learning systems systematically embed biases, while research institutes funded by mega-corporations are popping up to warn us of AI’s destructive potential. Writers of literary fiction and thoughtful film directors are discovering its metaphorical potential for exploring the subtleties of emotional life. The current exhibition at the Barbican, AI: More than Human, easily slots into a contemporary cultural landscape, which more or less has accepted that such a thing as artificial intelligence exists – or that someday soon it will.
The exhibition snakes its way round the edges of the building in the Barbican’s Curve Gallery, a layout that lends a strong sense of narrative trajectory on both chronological and conceptual levels in guiding the visitors’ experience. Keeping this in mind, the many artefacts and historical timelines lend the impression of an objectivity that should be taken with a grain of salt. Curated by Suzanne Livingston and Maholo Uchida, it is divided into four areas, though in practice this division spills out beyond the confines of the gallery space itself as the exhibition takes part in the Life: Rewired season at the Barbican. The exhibition, then, is only part of a number of installations throughout the building and a number of related talks and events have taken place over the months the season has lasted. The exhibition space is composed of a mix of artworks, historical artefacts and media projections, as well as certain elements that, at least to this reviewer, present thinly disguised corporate and government propaganda – making it ambiguous whether this undermines or reinforces the progressive narrative of AI.
The opening section, ‘The Dream of AI’, intends to show how AI, perhaps the defining speculative technology post-WWII, has roots in the supernatural imagination of the past. Creatures from superstition, religion and pre-scientific alchemy are paraded, and the Jewish golem is especially prominent from the beginning as the subject of It, a sound installation by Kode9 (Steve Goodman). A copy of Gustav Meyrink’s novel Der Golem (1915) and a projection of Paul Wegener’s third Golem film (1920; the only survivor of the trilogy) reminds us that this golem is, in fact, borne of modernity and war. In the exhibition catalogue, Goodman cites Norbert Wiener’s odd God and Golem, Inc. (1964) and Stanisław Lem’s’ ‘Golem XIV’ (1981) as inspirations for It. The sonic essay, though, is delivered via directed sound projection which makes it both difficult to make sense of and somewhat unsuited for the large number of visitors which pass through, as only a few people can listen in at a time.
The section is otherwise littered with pre-scientific detritus – cod-alchemical stills and alembics, Kabbalistic recipes and a prominent display of Japanese kami, the spirits that populate Shintoism. The kami are intended to express how the Japanese acceptance of robots connects to the belief in spirits that inhabit what we in the West otherwise think of as inanimate. A plaque claiming that the I Ching used binary numbers to represent the world hangs over a reproduction of Gottfred Leibniz’s calculator. Leibniz was indeed inspired by the I Ching when he developed binary numbers – but his calculator was not, and the process of signification that connects the binary numbers with Boolean logic and electric switches would not be realised for several hundred years. This magical genealogy strikes me as a kind of revisionism that prioritises aesthetic similarity and poetic allusion over historical accuracy. In weaving a narrative that emphasises a formal continuity over epistemic breaks, one wonders in which direction the revisionism goes: was the golem an automaton, or is the quest for AI a sorcerer’s search for the right incantation? To view the golem as artificial is an anachronism: it was supernatural, created by an occult appropriation of divine powers – matter given form through invocation; what it is not is a ghost in the machine. Similarly, the word intelligence is easily misinterpreted or even abused (as the quip goes, ‘intelligence’ is whatever an intelligence test tests for). From one point of view, the golem is intelligent because it has language, but in a deeper sense, it is not, for as a creature that takes every word literally, it lacks understanding. In this, as Wiener observed, the golem resembles the computer. The lack of understanding, as J.R. Searle observed, is why the computer is not intelligent.
The next section, ‘Mind Machines’, features many interesting pieces, including a part model of Charles Babbage’s Analytical Engine. A letter written by Ada Lovelace concerns her wish to create a calculus of the mind. Replica models of the Bletchley Park Bombe and the Enigma then leap a gap of a hundred years, obscuring the many innovations that took place in between Babbage and Alan Turing. As important as Turing is to our current understanding of AI, he did not believe his general-purpose computer was intelligent in itself, but that it was capable of simulating it. Babbage’s Difference Engine was created to automate error-prone calculation of tables, not to automate the mind. Alongside these objects is a densely annotated timeline on a series of touchscreens that intertwines computing history and speculative imaginations, beginning in 1818 with Mary Shelley’s Frankenstein.
Suzanne Livingston’s framing essay in the catalogue gives the impression that the golem, which she grants is legendary, was somehow also real: ‘Ultimately, the Golem destroyed its original maker before being re-birthed in culture to live in stories, cartoons, and film again and again’ (14). Livingston proudly carries her badge as co-founder of the Ccru, a para-academic group formed at the University of Warwick in the late 1990s. In the exhibition’s prominent featuring of the golem, Kabbala and alchemical homunculi we sense the influence of Nick Land’s numerological ramblings and concept of ’hyperstition’, a blend of superstition, hyperbole and the hyperreal which signifies those cultural invocations which, like magical sigils, brings the imaginary into being.
Artificial intelligence was a recurring topic in the writings of the Ccru, not least for its accelerationist potential for speeding up capitalism. I suspect that the essay by Livingston and Mark Fisher, which refers to a Binomic Institute and an AI assistant ‘familiar’ was originally intended for the Ccru-authored column in Mute Magazine (1993-5). The many toy robots, chess- and go-playing computers and chatbot simulations on display should perhaps be read according to the accelerationist narrative. Google is heavily involved with the prominent display surrounding AlphaGo’s victory over Lee Sedol, and DeepMind’s Demis Hassabis has an essay in the catalogue with European Go-champion Fan Hui; the many replicas on display are on loan from IBM as well as the Watson platform used on Jeopardy!. Several displays are interactive, letting them cast their spell over visitors marvelling over the grotesque dreamscapes built on deep learning image recognition platforms. Perhaps AI is a sign of folly, or maybe the allure of creating an intelligence goes deeper than merely making money. Reportedly, Google has invested $2b in DeepMind, with limited returns. But the cynic in me suspects this image of AI, presented as sf in the tradition of alchemy and magic, is as much PR as it is trivial. Computers were constructed not to simulate people, but to replace their jobs, whether in government or corporations. These friendly robot dogs represent the cheerful façade of the AI imaginary, tin cans with googly eyes; more k-pop than k-punk.
The following section, ‘Data Worlds’, brings us up to date in the contemporary age, where the black box reigns in machine and deep learning. But the name also implies the nature of a world in which large data sets are not only widely available but also intentionally generated to drive consumer and citizen behaviour. Neural networks and machine learning are possible only with fast computers and the availability of vast amounts of data. The question is for what purpose to use it. In this section, we see both uses and abuses of contemporary AI systems built on machine and deep learning algorithms. A small screen presents a cheery cartoon demonstrating the iniquities of the controversial Chinese social credits programme. Presented without commentary, we are left to make our own judgement about it, though the catalogue has an essay that describes the privacy concerns surrounding increased data collection in China. Will the AI revolution also mean political upheaval, or is it a force for surveillance, oppression and conservatism? Or is it merely another circuit round the hype cycle? The most conceptually interesting installation is Joy Buolamwini’s exploration of bias in facial recognition systems. In the accompanying essay, she describes how her face could not be recognised by a machine learning system she was working with at the MIT Media Lab – at least not until she put on a white mask. The algorithm she used was trained exclusively on white faces, incapable of recognising the Other.
We encounter the mask again at the very end of the exhibition, in the ‘Endless Evolution’ section. ‘Vespers II’, from Neri Oxman’s Mediated Matter Group at MIT, is a series of death masks augmented through computation and synthetic biology. Strikingly alien, these colourful, shimmering masks are insect-like, yet built on human templates. It is unclear how, exactly, the process is carried out nor how this relates to the exhibition’s subject matter, but the resulting objects are indeed beautiful.
The mask, however, now seems a symbol that unifies the exhibition’s sprawling narrative. In the first section is an Egyptian death mask taken from Sigmund Freud’s collection, juxtaposed with a presentation of the uncanny valley. Why this fascination, however subtle, with masks in an exhibition on artificial intelligence? Again, we see the ghostly imprint of the Ccru, with its fascination with Freudian thanatos. I am reminded of the persona in Roman politics, and the prosopon in Greek theatre. The Turing test is as much about theatre and performance as it is about technology; the mimic is impossible without mimesis, and underneath lies the premise of art and literature, for all narrative requires the inanimate to speak. Artificial life surrounds us every day in the stories we tell, and the exhibition claims as much, if not in as many words.
Chloe Wood et al, eds. 2019. AI: More than Human. London: Barbican International Enterprises.